Add pydocstyle to the ruff rules (#52136)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Signed-off-by: Harry Mellor <hej.mellor@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
@@ -1,8 +1,7 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
"""
|
||||
Benchmark for FlashInfer fused collective operations vs standard operations.
|
||||
"""Benchmark for FlashInfer fused collective operations vs standard operations.
|
||||
|
||||
This benchmark compares:
|
||||
1. FlashInfer's allreduce_fusion with trtllm backend
|
||||
@@ -446,9 +445,15 @@ def run_benchmarks(
|
||||
"""Run all benchmarks for given configuration.
|
||||
|
||||
Args:
|
||||
num_tokens: Number of tokens in the benchmark input.
|
||||
hidden_dim: Hidden dimension of the benchmark input.
|
||||
dtype: dtype of the benchmark input.
|
||||
use_residual: Whether to include a residual add in the fused op.
|
||||
no_oneshot: Skip the one-shot allreduce variants.
|
||||
allreduce_params: Shared parameters for FlashInfer fused allreduce.
|
||||
workspaces: Dict mapping backend name ("trtllm", "mnnvl") to workspace.
|
||||
quant_modes: Set of quantization modes: "none", "fp8", "fp4".
|
||||
|
||||
"""
|
||||
(
|
||||
input_tensor,
|
||||
|
||||
Reference in New Issue
Block a user